The variational method is an approximation technique used to estimate ground state energies in quantum systems.The variational method is a powerful approximation technique widely used in physics and applied mathematics. Its essence is to minimize or maximize a functional — a rule that assigns a number to a function. In quantum mechanics, the functional is usually the average energy. You set up a trial wave function with adjustable parameters and find those parameter values that minimize the energy. According to the variational principle, the obtained energy is never lower than the exact ground state, so by deepening the trial function (increasing the number of parameters), one can approach the truth arbitrarily closely. A similar approach is used in elasticity theory, hydrodynamics, and machine learning, where variational autoencoders learn to compress and reconstruct data.
The roots of the method go back to the 17th century — the brachistochrone problem solved by Johann Bernoulli. Systematic development of the calculus of variations was achieved in the works of Leonhard Euler and Joseph-Louis Lagrange. Walter Ritz in 1908 published a method of trial functions for solving boundary value problems, and in 1928 applied it to the quantum-mechanical helium atom, obtaining energy values close to experimental ones. Later, the method was generalized into a powerful computational tool with the advent of computers. In statistical physics, the variational approach was developed by Richard Feynman.
For a quantum system, a trial wave function ψ(α; x) with variational parameters α = (α1, α2, ...) is set up. The functional E(α) = ∫ ψ* H ψ dτ / ∫ |ψ|^2 dτ is calculated, where H is the Hamiltonian. Then E(α) is minimized: ∂E/∂α_i = 0. The solution gives optimal α and an energy estimate E0 ≤ E_exact. Convergence is guaranteed by the completeness theorem: if the trial family is dense in the required functional space, then the limit of E_min coincides with the exact energy. In variational Bayesian inference, the Kullback-Leibler divergence between the approximation and the true posterior distribution is minimized.
Quantum chemists calculate molecular shapes this way, and engineers determine the optimal rocket trajectory with minimal fuel consumption.
A path worn across a lawn by people is often the shortest route — a natural manifestation of the variational principle: the trajectory minimizes effort.

















